Skip to content

Get found in AI search

Get found in AI search

See whether the AI engines recommend your brand, every week, with the answers to prove it.

See whether the AI engines recommend your brand, every week, with the answers to prove it.

Enter a brand and a category
Enter a brand and a category

10 questions on one engine, free, no email, results in about a minute.

10 questions on one engine, free, no email, results in about a minute.

  • Perplexity
  • Google AI Overviews
  • Google AI Mode
  • ChatGPT
  • Gemini
  • Claude
  • Grok
  • DeepSeek
  • Perplexity
  • AI Overviews
  • AI Mode
  • ChatGPT
  • Gemini
  • Claude
  • Grok
  • DeepSeek

One workspace per brand, one number per week

One number per week

example

Before

"A client asked whether we come up in ChatGPT. I typed the question in on the call, we weren't in the answer, and then she asked if that was normal. I had no idea."

Dana Whitcombe, Marketing lead, Marrow & Co, March

After

"Same client, last Tuesday. I said 52 percent, plus or minus 14, and that it was 31 in March. She asked what the plus or minus was doing there, so I opened the methodology page and we read it together for about two minutes. Slightly awkward, and better than the version where I say a number and she nods."

Dana Whitcombe, Marketing lead, Marrow & Co, September

Marrow & CoHarbourlineQuillstackOstreyFennelworksTALLOW

Used by 140 brands and 38 agencies.

Where you stand

How often your brand comes up when people ask AI for a recommendation in your category. Per engine, refreshed every Monday, with the margin of error printed beside it.

Who comes up instead

Your competitors on the same questions, and the brands you never thought to name. Keylight reads every answer and names who is winning.

brand

appearance

margin

week

Jasper

34% ±18

+7

Copy.ai

41% ±17

+2

Writesonic

9% ±11

−1

Discovered in answers, not tracked

Anyword

18% ±15

Track

Rytr

14% ±13

Track

Surfer SEO

8% ±10

Track

Pages cited on questions Jasper was missing from

g2.com /categories/ai-writing-assistant

1,204

zapier.com /blog/best-ai-writing-generator

851

reddit.com /r/marketing/…ai_copywriting

648

techradar.com /best/ai-writer

402

Questions Jasper was missing from · 41 of 487

Which AI tool writes product descriptions for Shopify?
Best free AI copywriting tool for small business?
AI writing tool for real estate listings?
All 41

Which pages win

The websites the engines quote when they answer questions in your category, and which pages get cited on the questions you are missing from.

The questions you are missing from

Listed one by one, so you know which questions to look at.

Keylight shows you what is winning. It does not tell you what to do about it.

"A client asked where the 31 percent came from. I opened the answer and read out the sentence Perplexity had written about them. Took me two clicks to find it, which is two more than I want on a call, but it was there."

"A client asked where the 31 percent came from. I opened the answer and read out the sentence Perplexity had written about them. Took me two clicks to find it, which is two more than I want on a call, but it was there."

Priya Natarajan, Head of SEO, Fieldnote Digital

Why to believe the number

This is what lets the score go in a client report.

34% ±18

Every score carries its margin of error

Every score carries its margin of error

Printed beside the number in the email and the PDF, never in a tooltip.

Printed beside the number, in the email and in the PDF.

No movement is reported until the results show a real change

No movement is reported until the results show a real change

A number that could have moved by chance is not reported as a change. Keylight judges each change from the answers behind the two saved results.

Judged from the two saved results.

"For teams that need brand voice consistency, Jasper is a strong choice…"

The raw answers sit beside every chart

The raw answers sit beside every chart

Read what the engine actually said, with the brand marked in the text.

trykeylight.com/methodology

The method is a public page

The method is a public page

Anyone can read how a score is produced without logging in, including your client.

A number you can put in front of a client without being caught out.

A number you can put in front of a client without being caught out.

Pricing

Work out your weekly capacity
Engines to check
200 · about 8 topics
200 questions on 4 engines is 800 checks a week. Grok and Claude are capped at 10% of capacity each; the calculator shows the cap when you turn either on.
Weekly capacity needed800Business covers it, with 200 spare. $149 a month.
Work out your weekly capacity
Engines to check
200 · about 8 topics
200 questions on 4 engines is 800 checks a week. Grok and Claude are capped at 10% of capacity each; the calculator shows the cap when you turn either on.
Weekly capacity needed800Business covers it, with 200 spare. $149 a month.
All eight engines are included in every plan.
Solo
$79a month
  • 500 questions tracked weekly
  • 1 workspace
  • 5 seats
  • All eight engines
Start 14 days free
Card required. Cancel any time before day 15.Trial runs at 100 checks a week.
Businesscommon choice
$149a month
  • 1,000 questions tracked weekly
  • 5 workspaces
  • 20 seats
  • All eight engines
  • Extra capacity, $100 per 1,000
Start 14 days free
Card required. Cancel any time before day 15.Trial runs at 100 checks a week.
Agency
$449a month
  • 4,000 questions tracked weekly
  • 15 workspaces
  • Unlimited seats
  • White label and custom domain
  • Extra capacity, $100 per 1,000
Start 14 days free
Card required. Cancel any time before day 15.Trial runs at 100 checks a week.
Agency Plus
$899a month
  • 10,000 questions tracked weekly
  • Unlimited workspaces
  • Unlimited seats
  • White label and custom domain
  • Extra capacity, $100 per 1,000
Start 14 days free
Card required. Cancel any time before day 15.Trial runs at 100 checks a week.
Enterprise, for more than 10,000 tracked weekly or a contract. Talk to usGrok and Claude are each limited to 10% of your weekly capacity. The other six are unlimited.
All eight engines are included in every plan.
Businesscommon choice
$149a month
  • 1,000 questions tracked weekly
  • 5 workspaces
  • 20 seats
  • All eight engines
  • Extra capacity, $100 per 1,000
Start 14 days free
Card required. Cancel any time before day 15.Trial runs at 100 checks a week.
Solo
$79a month
  • 500 questions tracked weekly
  • 1 workspace
  • 5 seats
  • All eight engines
Start 14 days free
Card required. Cancel any time before day 15.Trial runs at 100 checks a week.
Agency
$449a month
  • 4,000 questions tracked weekly
  • 15 workspaces
  • Unlimited seats
  • White label and custom domain
  • Extra capacity, $100 per 1,000
Start 14 days free
Card required. Cancel any time before day 15.Trial runs at 100 checks a week.
Agency Plus
$899a month
  • 10,000 questions tracked weekly
  • Unlimited workspaces
  • Unlimited seats
  • White label and custom domain
  • Extra capacity, $100 per 1,000
Start 14 days free
Card required. Cancel any time before day 15.Trial runs at 100 checks a week.
Enterprise, for more than 10,000 tracked weekly or a contract. Talk to usGrok and Claude are each limited to 10% of your weekly capacity. The other six are unlimited.
All eight engines are included in every plan.
Solo
$79a month
  • 500 questions tracked weekly
  • 1 workspace
  • 5 seats
  • All eight engines
Start 14 days free
Card required. Cancel any time before day 15.Trial runs at 100 checks a week.
Businesscommon choice
$149a month
  • 1,000 questions tracked weekly
  • 5 workspaces
  • 20 seats
  • All eight engines
  • Extra capacity, $100 per 1,000
Start 14 days free
Card required. Cancel any time before day 15.Trial runs at 100 checks a week.
Agency
$449a month
  • 4,000 questions tracked weekly
  • 15 workspaces
  • Unlimited seats
  • White label and custom domain
  • Extra capacity, $100 per 1,000
Start 14 days free
Card required. Cancel any time before day 15.Trial runs at 100 checks a week.
Agency Plus
$899a month
  • 10,000 questions tracked weekly
  • Unlimited workspaces
  • Unlimited seats
  • White label and custom domain
  • Extra capacity, $100 per 1,000
Start 14 days free
Card required. Cancel any time before day 15.Trial runs at 100 checks a week.
Enterprise, for more than 10,000 tracked weekly or a contract. Talk to usGrok and Claude are each limited to 10% of your weekly capacity. The other six are unlimited.
CompareSoloBusinessAgencyAgency Plus
Questions tracked weekly5001,0004,00010,000
Workspaces1515Unlimited
Seats520UnlimitedUnlimited
Competitors per workspace3333
EnginesAll eightAll eightAll eightAll eight
Grok and Claude fair-use limit10% each10% each10% each10% each
Extra capacity, $100 per 1,000NoYesYesYes
Weekly email per workspaceYesYesYesYes
Monthly PDF reportBrandedBrandedBranded or unbrandedBranded or unbranded
White label and custom domainNoNoYesYes
Annual, two months free$790 a year$1,490 a year$4,490 a year$8,990 a year

"For about a year I checked ChatGPT by hand on Friday afternoons. Four questions and a screenshot, and I called that research. First month here the score did almost nothing and the product printed within noise next to it instead of drawing me a line going up. I have paid for software that would have drawn the line. Setup was fiddly, I had to name topics before it would run anything and I only had five."

Tom EverleighFounder, Harbourline Software

"Same G2 page quoted 40 times on questions we missed. Can't fix that, but I've stopped guessing."

Ines FerrandoGrowth, Quillstack

"It's a line on the invoice now, $250 a month per client for the AI visibility section. Nine clients on it, the plan holds fifteen, so I have room. The second client covered the $449 and the rest is margin. Only gripe is that the run lands on Monday, which is the day I'm writing reports."

Rob CastellanOwner, Castellan Creative

"Our founder was convinced one particular rival was all over AI answers. They came out at 12, we were at 19, and the two margins overlapped enough that the gap might be nothing at all. That killed an argument we'd been having every fortnight. I wanted a cleaner answer than that and I didn't get one."

Dana WhitcombeMarketing lead, Marrow & Co

"The PDF drops into our monthly report as its own section, with our logo on it and the raw answers printed after the chart, so if a client wants to check the score against what the engine actually said, it's on the next page. Two of mine now ask for it before I've sent it. For one client I use the unbranded version, because he doesn't want anyone's logo in an appendix."

Sam OkaforAccount director, Northlight Partners

"For about a year I checked ChatGPT by hand on Friday afternoons. Four questions and a screenshot, and I called that research. First month here the score did almost nothing and the product printed within noise next to it instead of drawing me a line going up. I have paid for software that would have drawn the line. Setup was fiddly, I had to name topics before it would run anything and I only had five."

Tom EverleighFounder, Harbourline Software

"Same G2 page quoted 40 times on questions we missed. Can't fix that, but I've stopped guessing."

Ines FerrandoGrowth, Quillstack

"It's a line on the invoice now, $250 a month per client for the AI visibility section. Nine clients on it, the plan holds fifteen, so I have room. The second client covered the $449 and the rest is margin. Only gripe is that the run lands on Monday, which is the day I'm writing reports."

Rob CastellanOwner, Castellan Creative

"Our founder was convinced one particular rival was all over AI answers. They came out at 12, we were at 19, and the two margins overlapped enough that the gap might be nothing at all. That killed an argument we'd been having every fortnight. I wanted a cleaner answer than that and I didn't get one."

Dana WhitcombeMarketing lead, Marrow & Co

"The PDF drops into our monthly report as its own section, with our logo on it and the raw answers printed after the chart, so if a client wants to check the score against what the engine actually said, it's on the next page. Two of mine now ask for it before I've sent it. For one client I use the unbranded version, because he doesn't want anyone's logo in an appendix."

Sam OkaforAccount director, Northlight Partners

Running this for clients?

One workspace per client, one shared pot of capacity, a white-labelled PDF each month and your own web address on the dashboard.

For agencies

Questions

How often does it check?

Every Monday. Each topic in a workspace becomes about 25 questions, and each is asked once a week on the engines you chose. Your first results arrive the Monday after you set up.

Which engines does it check?

Eight: Perplexity, Google AI Overviews, Google AI Mode, ChatGPT, Gemini, Claude, Grok and DeepSeek. All eight are in every plan. Grok and Claude are each limited to 10% of your weekly capacity. DeepSeek does not search the web, and every DeepSeek answer says so.

What does the margin of error mean?

It is the range the real figure probably sits in. 34% ±18 means the brand came up in 34% of answers, and the true rate is likely between 16% and 52%. More questions narrow the range, which is why each topic asks about 25.

Do I install anything?

No. Keylight is a website you log into. There is nothing to add to your own site and nothing to connect.

What does the free report card cost?

Nothing. It asks 10 questions on one engine, needs no email, and the result arrives in about a minute. Each domain can be checked once a week.

Can I track competitors?

Yes. Each workspace tracks your brand and up to 3 competitors on the same questions, and lists the brands that came up that nobody typed in.

What happens when I hit my capacity?

Questions past your weekly capacity are not checked that week, and nothing extra is charged. On Business and above you can add capacity from inside the product, $100 a month for 1,000 more a week.

How do I cancel?

From the Billing page, which opens your account with Paddle, who handle payments for us. Checks stop at the end of the period you paid for.

Find out in about a minute

Run the free report card